openclaw - 💡(How to fix) Fix [Bug] fetchWithSsrFGuard consistently times out for xiaomi-coding provider (2min timeout, no response)

Summary

Fetch timeout repeatedly occurs when querying xiaomi-coding provider (https://token-plan-cn.xiaomimimo.com/v1/chat/completions).

Environment

• OS: Windows 11 (10.0.26200)
• OpenClaw version: 2026.5.10-beta.2
• Node.js: v24.15.0

Steps to reproduce

The xiaomi-coding provider is configured as a standard OpenAI-compatible endpoint:

{
  "xiaomi-coding": {
    "baseUrl": "https://token-plan-cn.xiaomimimo.com/v1",
    "api": "openai-completions",
    "models": [
      { "id": "mimo-v2.5", "name": "mimo-v2.5", "reasoning": true, "contextWindow": 1048576, "maxTokens": 32000 },
      { "id": "mimo-v2.5-pro", "name": "mimo-v2.5-pro", "reasoning": true, "contextWindow": 1048576, "maxTokens": 32000 }
    ]
  }
}

It is used as a fallback model for the adai agent and as a standalone model for others.

Actual behavior

The fetch request to https://token-plan-cn.xiaomimimo.com/v1/chat/completions consistently times out after 120 seconds (the internal fetchWithSsrFGuard default timeout). This happens across multiple hours and days (observed 21:05, 21:44, 23:17 on 2026-05-11, and earlier).

Sample log:

[fetch-timeout] fetch timeout after 120000ms (elapsed 120024ms)
operation=fetchWithSsrFGuard
url=https://token-plan-cn.xiaomimimo.com/v1/chat/completions

Expected behavior

The request should complete (success or error) within a reasonable time, or the timeout should be configurable per-provider.

Questions

1. Is this a network/DNS issue on the OpenClaw side, or is the upstream server (token-plan-cn.xiaomimimo.com) simply unresponsive?
2. Can the fetch timeout be made configurable per provider? A global 120s is too long to wait when the upstream is dead, but may be too short for large context requests.
3. Are there any known issues with fetchWithSsrFGuard and long-lived connections or connection pooling?

Additional context

• Other providers (Nvidia, LongCat, FreeModel) work fine — only xiaomi-coding has this issue
• Direct curl test from the same machine may shed light on whether it is a network-level issue (TBD)
• The issue may be reproducible with any OpenAI-compatible provider that has slow/unstable endpoints